EP2332301B1 - Method and device for quantizing likelihood quotients - Google Patents
Method and device for quantizing likelihood quotients Download PDFInfo
- Publication number
- EP2332301B1 EP2332301B1 EP09764705A EP09764705A EP2332301B1 EP 2332301 B1 EP2332301 B1 EP 2332301B1 EP 09764705 A EP09764705 A EP 09764705A EP 09764705 A EP09764705 A EP 09764705A EP 2332301 B1 EP2332301 B1 EP 2332301B1
- Authority
- EP
- European Patent Office
- Prior art keywords
- quantizing
- values
- likelihood
- levels
- quantization
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims description 45
- 238000013139 quantization Methods 0.000 description 79
- 230000009467 reduction Effects 0.000 description 12
- 238000005457 optimization Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 6
- 230000002902 bimodal effect Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 230000007423 decrease Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 230000014509 gene expression Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000013144 data compression Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 240000006829 Ficus sundaica Species 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000002945 steepest descent method Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/06—Dc level restoring means; Bias distortion correction ; Decision circuits providing symbol by symbol detection
- H04L25/067—Dc level restoring means; Bias distortion correction ; Decision circuits providing symbol by symbol detection providing soft decisions, i.e. decisions together with an estimate of reliability
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0045—Arrangements at the receiver end
- H04L1/0047—Decoding adapted to other signal detection operation
- H04L1/005—Iterative decoding, including iteration between signal detection and decoding operation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0064—Concatenated codes
- H04L1/0066—Parallel concatenated codes
Definitions
- the invention relates to a method for signal processing for quantizing likelihood quotients of binary random variables for transmission between two or more signal processing units.
- the invention relates to a quantization device in signal processing for quantizing likelihood quotients of binary random variables for transmission between two or more signal processing units.
- interlinked signal processing units exchange probabilities to perform certain tasks in a signal processing chain or an iterative one Signal processing.
- An example of such a type of signal processing is a decoding of turbo-code, which is used in mobile networks of the third generation (eg UMTS).
- turbo-code which is used in mobile networks of the third generation (eg UMTS).
- two decoders each generate a hypothesis regarding a received payload bit pattern.
- probabilities so-called likelihood values are derived and used.
- the decoders exchange the likelihood values and each generate a new hypothesis taking into account the likelihood values of the other decoder. This process is repeated iteratively until both decoders provide the same hypothesis for the payload bit pattern.
- P X ( x ) describes the probability that the random variable X assumes the value x .
- Random variables such as X or L ( X ) are here denoted by capital letters and their realization x , l by lowercase letters.
- log is meant the natural logarithm.
- the log likelihood quotient L ( X ) is also referred to below as the L value or LLR (Log Likelihood Ratio) of the random variable X.
- the range of values of L ( X ) extends from - ⁇ to + ⁇ .
- describes the reliability of this decision.
- the distribution of the L values in the value range ⁇ ⁇ or the density of the L values depends on the noise occurring in a channel used for transmitting x values.
- a bimodal Gaussian distribution of L values follows.
- AWGN additive white Gaussian noise
- flat Rayleigh channels and generally also approximately to large data block lengths.
- L values must be transmitted with finite precision between different signal processing units for digital processing.
- the requisite quantization step causes unavoidable quantization noise, which can be equated with information loss in the processing system.
- scalar quantization maps continuous L values to a limited set of R discrete values r i , which are referred to as quantization values, quantization levels, or reconstruction levels.
- An L value in an interval [ d i , d i +1 ) is mapped to the corresponding reconstruction level r i .
- the terms reconstruction level and decision level are used.
- the information loss is reduced at the cost of additional data transfer and processing overhead. For this reason, the number R of reconstruction levels r i is limited. A further reduction of the information loss is possible by a suitable determination of the reconstruction levels r i and the decision levels d i . On the other hand, adequate quantization allows for an advantageous reduction of the data rate.
- a well-known non-uniform quantization method for scalar quantities is the Lloyd-Max quantization after S. Lloyd, "Least squares quantization in PCM,” (unpublished Bell Lab., Techn., 1957) IEEE Trans. Inf. Theory, vol. IT-28, pp. 129-137, 1982 and J. Max, Quantizing for Minimum Distortion, IEEE Trans. Inf. Theory, vol. 6, pp. 7-12, 1960 .
- MSE Mean Square Error
- the decision levels d i and the reconstruction levels r i are adjusted in several iterative steps. For log likelihood ratios, this approach disadvantageously results in many levels being used for large value L values to keep the mean squared error small, although a change in reliability is already very small at these L values.
- the object of the invention is therefore to avoid disadvantages of the prior art and to optimize a mutual coordination of data rate and accuracy in a quantization of likelihood quotients for transmission.
- the object is achieved by a method according to claim 1. Furthermore, the object is achieved with a quantization device according to claim 6.
- the mutual information between the binary variables and the quantized likelihood quotients can be represented as a function of the reconstruction and decision levels for a given probability distribution or probability density.
- the mutual information is maximized before quantizing the likelihood quotients by varying the reconstruction and decision levels. For this purpose, for example, a steepest descent method or another suitable numerical optimization method can be used.
- the optimized levels are then used to quantize the likelihood quotients.
- Such a direct optimization of the reconstruction and decision levels for the maximum mutual information according to the method and the device according to the invention minimizes a loss of mutual information during quantization.
- a loss of mutual information When transmitting quantized likelihood quotients between signal processing units, therefore, advantageously less quantization noise occurs.
- the loss of mutual information is smaller and a signal processing process is performed much more accurately.
- a reduction of the data rate by fewer levels with the same loss of mutual information is possible, whereby signal processing is faster and with less effort operable.
- An advantageous embodiment of the method according to the invention in a signal processing for the quantization of likelihood quotients is achieved by detecting a variance of the likelihood quotients before a quantization for the determination of decision levels and reconstruction levels.
- the variance is useful, for example, in a bimodal Gaussian distribution, as occurs in AWGN channels or flat Rayleigh channels, for describing the probability distribution or probability density of the likelihood quotients.
- For a quantization it is thus possible to optimally determine reconstruction and decision levels adapted to a given variance of the likelihood quotients.
- the loss of Mutual information by quantization is minimized according to a given variance.
- a further preferred embodiment of the method according to the invention provides for adaptation of an approximation function as a function of the variance to determined reconstruction levels or decision levels in the case of different variances and determination of reconstruction levels and decision levels in the case of quantization by the approximation function.
- a corresponding approximation function is used in each case for one, several or all reconstruction and decision levels.
- a value table and / or an interpolation are used when determining reconstruction levels or decision levels. This is e.g. additionally or alternatively to an approximation function possible. In this way, determination of optimum levels in the sense of maximum mutual information can also be carried out quickly and without expensive optimization methods prior to quantization.
- entropy coding of reconstruction levels is carried out. Quantization with maximization of mutual information results in reconstruction levels that are used with different probability.
- the reconstruction levels are assigned level identifiers or bit patterns of different lengths. A prefix-free code is used and more frequent reconstruction levels receive a shorter level descriptor.
- a Huffman coding is used as entropy coding. As a result, a data compression is achieved with to be transmitted reconstruction levels. This leads to an effective reduction of a data rate when transmitting quantized likelihood quotients.
- An advantageous embodiment of the quantization device according to the invention in a signal processing for quantizing likelihood quotients is achieved in that approximation means for fitting an approximation function as a function of the variance to determined reconstruction levels or decision levels at different variances and a level generator for determining reconstruction levels and decision levels in a quantization with the approximation function are provided.
- approximation means for fitting an approximation function as a function of the variance to determined reconstruction levels or decision levels at different variances and a level generator for determining reconstruction levels and decision levels in a quantization with the approximation function are provided.
- a time-consuming and resource-intensive maximization of the mutual information by an optimization method is thereby avoided.
- Optimal levels depending on a given variance are determined quickly and easily.
- Mutual information or transinformation is a quantity known to the expert from information theory, which indicates the strength of the statistical relationship of two random variables.
- the mutual information 1 ( X; L ) between two variables x and l describes the mean reduction of the uncertainty of x if l is known and vice versa. This is for example in Thomas M. Cover, Joy A. Thomas, Elements of Information Theory 2nd Ed., John Wiley & Sons, 2006 described.
- the mutual information is used here to determine the mean loss of information of binary data when quantizing associated L values.
- the mutual information 1 ( X; L ) is between the binary variable X ⁇ ⁇ +1, -1 ⁇ and the associated quantized, discrete L value density and the unquantized one , continuous L value density calculable.
- x 1 + p ⁇ l
- x - 1 ⁇ dl
- ⁇ I q ⁇ r i 1 4 ⁇ ln ⁇ 2 ⁇ e - r i 1 + e r i ⁇ erf ⁇ l - ⁇ L 2 ⁇ ⁇ L ⁇
- the mutual information I q of the quantized L value density must be maximized.
- maximization of the mutual information I q is achieved by correspondingly adjusting the reconstruction levels r i and decision levels d i .
- Fig. 1 shows a schematic flow diagram of an embodiment.
- a steepest descent process or another suitable numerical optimization method for maximizing the mutual information I q rectangle 10.
- means are provided in a corresponding quantization device for determining reconstruction and decision levels by maximizing the mutual information.
- optimized reconstruction levels r i and decision levels d i result with respect to a minimum information loss ⁇ I.
- the quantization is compared by maximizing the mutual information (max MI quant.) With the direct quantization of the L values (LLR quant.).
- LLR quant. the direct quantization of the L values
- LLR quant. the direct quantization of the L values
- the reduction decreases with increasing number R of levels.
- the other diagram of the Fig. 3 compares the quantization by maximizing the mutual information (max MI quant.) with the soft-bit quantization (soft bit quant.).
- a suitable approximation function of the variance becomes ⁇ L 2 or the standard deviation ⁇ L to some values determined by maximizing the mutual information of a reconstruction level r i or decision levels d i for different variances ⁇ L 2 customized.
- the quantization device approximation means are available.
- f ⁇ L a ⁇ ⁇ L + b ⁇ ⁇ L + c ⁇ ⁇ L 2 used.
- Fig. 2 For example, for two reconstruction levels and two decision levels, correspondingly adapted functions f ( ⁇ L ) are shown as lines.
- suitable coefficients a, b and c are determined for a level.
- the variance ⁇ L 2 or the standard deviation ⁇ L of the L values is determined with a detection unit, rectangle 14. With the aid of the approximation function f ( ⁇ 2 ) and the corresponding coefficients a, b and c , then dependent on ⁇ L 2 or ⁇ L optimal reconstruction levels r i and decision levels d i fixed with respect to the exchange of information by a level generator rectangle 16. Finally carried quantization rectangle 18, the L values with use of these reconstruction levels r i and decision levels d i by a quantization unit.
- the method steps 14, 16 and 18 are re-performed, for example, in an iterative signal processing for each quantization of L values.
- Such a code is for example the expert from Thomas M. Cover, Joy A. Thomas, Elements of Information Theory 2nd Ed., John Wiley & Sons, 2006 known.
- a BCJR algorithm is used for decoding, with the Maximum-Log-MAP-Approximation (MAP: Maximum A posteriori Probability) yielding substantially the same results as the exact Log-MAP version.
- MAP Maximum A posteriori Probability
- the variance of the L values is adaptively determined at the MAP decoder output with the detection means.
- , the absolute L value of the code bits, on the variance ⁇ L 2 displayed.
- the level generator is used in the quantization device. The determined levels are thus optimized to maximize the mutual information depending on the variance.
- bit error rate (BER) of the individual decodes is represented as a function of the bit energy to noise power density ratio E b / N 0 .
- the ratio of the energy E b applied to an information bit to the spectral noise power density N 0 relates the signal-to-noise ratio to an information bit.
- E b / N 0 With increasing noise, ie decreasing E b / N 0 usually increases the bit error rate.
- the bit error rates for v arious R and E b / N 0 when using the quantization are shown (max.
- Quantisation MI Quantisation MI with maximization of mutual information.
- the decrease of the bit error rate with increasing R can be clearly seen.
- the non quantized representation of the a priori information is used as reference (no quant.).
- the lower graph shows bit error rates for corresponding R and E b / N 0 using direct Lloyd-Max quantization for the L values. This quantization generally results in a higher bit error rate than the max MI quantization.
- an entropy coding, rectangle 20, of the reconstruction levels is additionally performed by means of an entropy coder of the quantization device.
- Entropy encodings are the expert from, for example A. Gersho, RM Gray, "Vector Quantization and Signal Compression,” Boston, MA, Kluwer, 1992 and Thomas M. Cover, Joy A. Thomas, Elements of Information Theory 2nd Ed., John Wiley & Sons, 2006 known.
- Quantization to maximize mutual information as well as soft-bit quantization provides reconstruction levels that are used with different probability. The areas under the bimodal Gaussian density for each reconstruction level are different.
- the reconstruction levels are assigned level identifiers or bit patterns of different lengths. A prefix-free code is used and more frequently used reconstruction levels receive a shorter level indicator. In this way, a data compression and thus an effective reduction of the data rate is achieved.
- R ( L q ) becomes a function of the variance ⁇ L 2 for different R (number of reconstruction levels) shown by solid lines.
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Power Engineering (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Complex Calculations (AREA)
- Error Detection And Correction (AREA)
Description
Die Erfindung betrifft ein Verfahren bei einer Signalverarbeitung zur Quantisierung von Likelihood-Quotienten binärer Zufallsvariablen für eine Übermittlung zwischen zwei oder mehreren Signalverarbeitungseinheiten.The invention relates to a method for signal processing for quantizing likelihood quotients of binary random variables for transmission between two or more signal processing units.
Weiterhin betrifft die Erfindung eine Quantisierungseinrichtung bei einer Signalverarbeitung zum Quantisierung von Likelihood-Quotienten binärer Zufallsvariablen für eine Übermittlung zwischen zwei oder mehreren Signalverarbeitungseinheiten.Furthermore, the invention relates to a quantization device in signal processing for quantizing likelihood quotients of binary random variables for transmission between two or more signal processing units.
Es sind viele Signalverarbeitungsvorgängen bekannt, bei denen miteinander verknüpfte Signalverarbeitungseinheiten Wahrscheinlichkeiten austauschen, um bestimmte Aufgaben in einer Signalverarbeitungskette durchzuführen oder eine iterative Signalverarbeitung zu ermöglichen. Ein Beispiel einer solchen Art von Signalverarbeitung ist eine Decodierung von Turbo-Code, welcher in Mobilfunknetzen der dritten Generation (z.B. UMTS) zum Einsatz kommt. In diesem Beispiel erzeugen zwei Decoder jeweils eine Hypothese bezüglich eines empfangenen Nutzbitmusters. Dazu werden Wahrscheinlichkeiten, sogenannte Likelihood-Werte hergeleitet und verwendet. Die Decoder tauschen die Likelihood-Werte aus und erzeugen jeweils eine neue Hypothese unter Berücksichtigung der Likelihood-Werte des anderen Decoders. Dieser Vorgang wird iterativ solange wiederholt, bis beide Decoder die gleiche Hypothese für das Nutzbitmuster liefern.Many signal processing operations are known in which interlinked signal processing units exchange probabilities to perform certain tasks in a signal processing chain or an iterative one Signal processing. An example of such a type of signal processing is a decoding of turbo-code, which is used in mobile networks of the third generation (eg UMTS). In this example, two decoders each generate a hypothesis regarding a received payload bit pattern. For this purpose, probabilities, so-called likelihood values are derived and used. The decoders exchange the likelihood values and each generate a new hypothesis taking into account the likelihood values of the other decoder. This process is repeated iteratively until both decoders provide the same hypothesis for the payload bit pattern.
Oft werden sogenannte Log-Likelihood-Quotienten erzeugt und als Wahrscheinlichkeitswerte zwischen verschiedenen Signalverarbeitungseinheiten ausgetauscht. Der Log-Likelihood-Quotient einer binären Zufallsvariable X ∈ {+1, -1} ist definiert als:
Dabei beschreibt PX (x) die Wahrscheinlichkeit, dass die Zufallsvariable X den Wert x annimmt. Zufallsvariablen wie X oder L(X) werden hier mit Großbuchstaben und ihre Realisation x, l mit Kleinbuchstaben bezeichnet. Unter "log" wird der natürliche Logarithmus verstanden. Der Log-Likelihood-Quotient L(X) wird im Folgenden auch als L-Wert oder LLR (Log-Likelihood-Ratio) der Zufallsvariable X bezeichnet. Der Wertebereich von L(X) erstreckt sich von -∞ bis +∞. Das Vorzeichen von L(X) ist die Entscheidung, ob x = +1 oder x = -1 ist. Diese Entscheidung wird auch Hard-Decision genannt. Der Betrag |L(X)| beschreibt die Zuverlässigkeit dieser Entscheidung.Here, P X ( x ) describes the probability that the random variable X assumes the value x . Random variables such as X or L ( X ) are here denoted by capital letters and their realization x , l by lowercase letters. By "log" is meant the natural logarithm. The log likelihood quotient L ( X ) is also referred to below as the L value or LLR (Log Likelihood Ratio) of the random variable X. The range of values of L ( X ) extends from -∞ to + ∞. The sign of L ( X ) is the decision as to whether x = +1 or x = -1. This decision is also called hard decision. The amount | L ( X ) | describes the reliability of this decision.
Die Verteilung der L-Werte in dem Wertebereich ±∞ bzw. die Dichte der L-Werte hängt von dem auftretenden Rauschen in einem zur Übermittlung von x-Werten verwendeten Kanal ab. Für einen gaußschen Kanal folgt eine bimodale Gauß-Verteilung der L-Werte.The distribution of the L values in the value range ± ∞ or the density of the L values depends on the noise occurring in a channel used for transmitting x values. For a Gaussian channel, a bimodal Gaussian distribution of L values follows.
Da für Mittelwert µ und Varianz σ2 die Beziehung
Die Annahme eines gaußschen Kanals gilt für AWGN- (Additives Weißes Gaußsches Rauschen) oder flache Rayleigh-Kanäle und im Allgemeinen auch näherungsweise für große Datenblocklängen.The assumption of a Gaussian channel applies to AWGN (additive white Gaussian noise) or flat Rayleigh channels, and generally also approximately to large data block lengths.
Die L-Werte müssen mit endlicher Genauigkeit zwischen verschiedenen Signalverarbeitungseinheiten zur digitalen Verarbeitung übermittelt werden. Der dazu notwendige Quantisierungsschritt verursacht unvermeidbares Quantisierungsrauschen, welches mit einem Informationsverlust in dem Verarbeitungssystem gleichgesetzt werden kann. Im Allgemeinen bildet eine skalare Quantisierung kontinuierliche L-Werte auf einen begrenzten Satz von R diskreten Werten ri ab, welche als Quantisierungswerte, Quantisierungslevel oder Rekonstruktionsniveaus bezeichnet werden. Zu diesem Zweck werden D (= R + 1) Schwellwerte oder Entscheidungsniveaus di in dem Wertebereich ±∞ der L-Werte festgelegt. Ein L-Wert in einem Intervall [di , d i+1) wird auf das entsprechende Rekonstruktionsniveau ri abgebildet. Im Folgenden werden die Bezeichnungen Rekonstruktionsniveau und Entscheidungsniveau verwendet.The L values must be transmitted with finite precision between different signal processing units for digital processing. The requisite quantization step causes unavoidable quantization noise, which can be equated with information loss in the processing system. In general, scalar quantization maps continuous L values to a limited set of R discrete values r i , which are referred to as quantization values, quantization levels, or reconstruction levels. For this purpose, D (= R + 1) threshold values or decision levels d i are set in the value range ± ∞ of the L values. An L value in an interval [ d i , d i +1 ) is mapped to the corresponding reconstruction level r i . In the following, the terms reconstruction level and decision level are used.
Durch ein Erhöhen der Anzahl R der Rekonstruktionsniveaus ri wird der Informationsverlust auf Kosten eines zusätzlichen Datentransfer- und Verarbeitungsaufwands reduziert. Aus diesem Grund ist die Anzahl R von Rekonstruktionsniveaus ri begrenzt. Eine weitere Reduzierung des Informationsverlustes ist durch eine geeignete Festlegung der Rekonstruktionsniveaus ri und der Entscheidungsniveaus di möglich. Andererseits ermöglicht eine adäquate Quantisierung eine vorteilhafte Reduktion der Datenrate.By increasing the number R of reconstruction levels r i , the information loss is reduced at the cost of additional data transfer and processing overhead. For this reason, the number R of reconstruction levels r i is limited. A further reduction of the information loss is possible by a suitable determination of the reconstruction levels r i and the decision levels d i . On the other hand, adequate quantization allows for an advantageous reduction of the data rate.
Bei einer bimodalen Gauß-Verteilung der L-Werte führt eine uniforme Quantisierung mit gleichmäßigen Intervallen [di, d i+1] und Rekonstruktionsniveaus ri in der Mitte dieser Intervalle zu erheblichen Fehlern bzw. einem hohen Quantisierungsrauschen. Viele verschiedene L-Werte werden auf Rekonstruktionsniveaus ri weniger Intervalle abgebildet, während andere Rekonstruktionsniveaus ri nur selten verwendet werden. Dabei werden insbesondere eine Wahrscheinlichkeitsverteilung der L-Werte und ein Informationsgehalt der L-Werte nicht berücksichtigt. Die Quantisierungsfehler setzten sich besonders bei einer iterativen Signalverarbeitung fort und wirken sich negativ auf den gesamten Signalverarbeitungsprozess aus.In a bimodal Gaussian distribution of the L values, uniform quantization with uniform intervals [ d i , d i +1 ] and reconstruction levels r i in the middle of these intervals leads to significant errors or high quantization noise. Many different L values are mapped to reconstruction levels r i less intervals, while other reconstruction levels r i are rarely used. In particular, a probability distribution of the L values and an information content of the L values are not taken into account. The quantization errors continue especially in an iterative signal processing and have a negative effect on the entire signal processing process.
Ein bekanntes nicht uniformes Quantisierungsverfahren für skalare Größen ist die Lloyd-Max-Quantisierung nach
In der Druckschrift
Es besteht jedoch weiterhin ein Bedarf nach einem optimaleren Quantisierungsverfahren von Log-Likelihood-Quotienten um Fehler in der Signalverarbeitung und den Arbeitsaufwand zu minimieren. Insbesondere ist der mittleren quadratischen Fehler zwischen kontinuierlichen und quantisierten Log-Likelihood-Quotienten für eine Bestimmung von Entscheidungsniveaus di und Rekonstruktionsniveaus ri kein optimales Kriterium.However, there remains a need for a more optimal quantization method of log likelihood ratios to minimize errors in signal processing and workload. In particular, the mean square error between continuous and quantized log-likelihood quotients is not an optimal criterion for determining decision levels d i and reconstruction levels r i .
Aufgabe der Erfindung ist es daher, Nachteile des Standes der Technik zu vermeiden und eine gegenseitige Abstimmung von Datenrate und Genauigkeit bei einer Quantisierung von Likelihood-Quotienten für eine Übermittlung zu optimieren.The object of the invention is therefore to avoid disadvantages of the prior art and to optimize a mutual coordination of data rate and accuracy in a quantization of likelihood quotients for transmission.
Erfindungsgemäß wird die Aufgabe mit einem Verfähren entsprechend Anspruch 1 gelöst. Weiterhin wird die Aufgabe mit einer Quantisierungseinrichtung entsprechend Anspruch 6 gelöst.According to the invention the object is achieved by a method according to
Die gegenseitige Information zwischen den binären Variablen und den quantisierten Likelihood-Quotienten lässt sich bei einer vorgegebenen Wahrscheinlichkeitsverteilung oder Wahrscheinlichkeitsdichte als Funktion der Rekonstruktions- und Entscheidungsniveaus darstellen. Erfindungsgemäß wird die gegenseitige Information vor einer Quantisierung der Likelihood-Quotienten durch Variieren der Rekonstruktions- und Entscheidungsniveaus maximiert. Dazu kann beispielsweise ein Verfahren des steilsten Abstiegs oder ein anderes geeignetes numerisches Optimierungsverfahren angewendet werden. Die optimierten Niveaus werden anschließend zum Quantisieren der Likelihood-Quotienten benutzt.The mutual information between the binary variables and the quantized likelihood quotients can be represented as a function of the reconstruction and decision levels for a given probability distribution or probability density. According to the invention, the mutual information is maximized before quantizing the likelihood quotients by varying the reconstruction and decision levels. For this purpose, for example, a steepest descent method or another suitable numerical optimization method can be used. The optimized levels are then used to quantize the likelihood quotients.
Durch eine solche direkte Optimierung der Rekonstruktions- und Entscheidungsniveaus bezüglich der maximalen gegenseitigen Information nach dem erfindungsgemäßen Verfahren und der erfindungsgemäßen Einrichtung wird ein Verlust an gegenseitiger Information bei der Quantisierung minimiert. Bei einer Übermittlung von quantisierten Likelihood-Quotienten zwischen Signalverarbeitungseinheiten tritt daher vorteilhaft weniger Quantisierungsrauschen auf. Bei gleichbleibender Datenrate einer Übermittlung ist der Verlust an gegenseitiger Information kleiner und ein Signalverarbeitungsprozess wird wesentlich genauer durchgeführt. Andererseits ist eine Reduzierung der Datenrate durch weniger Niveaus bei gleichem Verlust an gegenseitiger Information möglich, wodurch eine Signalverarbeitung schneller und mit weniger Aufwand betreibbar ist.Such a direct optimization of the reconstruction and decision levels for the maximum mutual information according to the method and the device according to the invention minimizes a loss of mutual information during quantization. When transmitting quantized likelihood quotients between signal processing units, therefore, advantageously less quantization noise occurs. At a constant data rate of transmission, the loss of mutual information is smaller and a signal processing process is performed much more accurately. On the other hand, a reduction of the data rate by fewer levels with the same loss of mutual information is possible, whereby signal processing is faster and with less effort operable.
Eine vorteilhafte Ausbildung des erfindungsgemäßen Verfahrens bei einer Signalverarbeitung zur Quantisierung von Likelihood-Quotienten wird durch ein Erfassen einer Varianz der Likelihood-Quotienten vor einer Quantisierung für das Ermitteln von Entscheidungsniveaus und Rekonstruktionsniveaus erzielt. Die Varianz eignet sich z.B. bei einer bimodalen Gaus-Verteilung, wie sie bei AWGN-Kanälen oder flachen Rayleigh-Kanäle auftritt, zur Beschreibung der Wahrscheinlichkeitsverteilung bzw. Wahrscheinlichkeitsdichte der Likelihood-Quotienten. Für eine Quantisierung lassen sich somit optimal an eine jeweils gegebene Varianz der Likelihood-Quotienten angepasste Rekonstruktions- und Entscheidungsniveaus ermitteln. Der Verlust an gegenseitiger Information durch eine Quantisierung wird jeweils entsprechend einer vorliegenden Varianz minimiert.An advantageous embodiment of the method according to the invention in a signal processing for the quantization of likelihood quotients is achieved by detecting a variance of the likelihood quotients before a quantization for the determination of decision levels and reconstruction levels. The variance is useful, for example, in a bimodal Gaussian distribution, as occurs in AWGN channels or flat Rayleigh channels, for describing the probability distribution or probability density of the likelihood quotients. For a quantization, it is thus possible to optimally determine reconstruction and decision levels adapted to a given variance of the likelihood quotients. The loss of Mutual information by quantization is minimized according to a given variance.
Eine weitere bevorzugte Ausgestaltung des erfindungsgemäßen Verfahrens sieht ein Anpassen einer Näherungsfunktion als Funktion der Varianz an ermittelte Rekonstruktionsniveaus oder Entscheidungsniveaus bei verschiedenen Varianzen und ein Ermitteln von Rekonstruktionsniveaus und Entscheidungsniveaus bei einer Quantisierung durch die Näherungsfunktion vor. Dabei wird beispielsweise für ein, mehrere oder alle Rekonstruktions- und Entscheidungsniveaus jeweils eine entsprechende Näherungsfunktion verwendet. Durch diese Maßnahme wird eine schnelle und unkomplizierte Bestimmung optimaler Niveaus in Abhängigkeit einer vorgegebenen Varianz ermöglicht. Eine zeitaufwendige und umständliche Maximierung der gegenseitigen Information durch ein geeignetes Optimierungsverfahren muss nicht vor jeder Quantisierung durchgeführt werden.A further preferred embodiment of the method according to the invention provides for adaptation of an approximation function as a function of the variance to determined reconstruction levels or decision levels in the case of different variances and determination of reconstruction levels and decision levels in the case of quantization by the approximation function. In this case, a corresponding approximation function is used in each case for one, several or all reconstruction and decision levels. By this measure, a fast and easy determination of optimal levels depending on a given variance is possible. A time-consuming and cumbersome maximization of the mutual information by a suitable optimization method need not be performed before each quantization.
Vorzugsweise werden bei einer Ausgestaltung des erfindungsgemäßen Verfahrens eine Wertetabelle und/oder eine Interpolation beim Ermitteln von Rekonstruktionsniveaus oder Entscheidungsniveaus verwendet. Dieses ist z.B. zusätzlich oder alternativ zu einer Näherungsfunktion möglich. Auf diese Weise wird ebenfalls eine Bestimmung optimaler Niveaus im Sinne maximaler gegenseitiger Information schnell und ohne aufwendige Optimierungsverfahren vor einer Quantisierung durchführbar.Preferably, in one embodiment of the method according to the invention, a value table and / or an interpolation are used when determining reconstruction levels or decision levels. This is e.g. additionally or alternatively to an approximation function possible. In this way, determination of optimum levels in the sense of maximum mutual information can also be carried out quickly and without expensive optimization methods prior to quantization.
Ferner wird in einer vorteilhaften Ausbildung des erfindungsgemäßen Verfahrens eine Entropiekodierung von Rekonstruktionsniveaus durchgeführt. Bei einer Quantisierung mit Maximierung der gegenseitigen Information ergeben sich Rekonstruktionsniveaus, welche mit unterschiedlicher Wahrscheinlichkeit benutzt werden. Durch eine Entropiekodierung werden den Rekonstruktionsniveaus Niveaubezeichner bzw. Bitmustern von unterschiedlicher Länge zugeordnet. Dabei wird ein präfixfreier Code verwendet und häufiger auftretende Rekonstruktionsniveaus erhalten einen kürzeren Niveaubezeichner. Als Entropiekodierung wird beispielsweise eine Huffman-Kodierung verwendet. Hierdurch wird eine Datenkompression bei zu übermittelnden Rekonstruktionsniveaus erreicht. Dies führt zu einer wirkungsvollen Reduzierung einer Datenrate bei einer Übermittlung von quantisierten Likelihood-Quotienten.Furthermore, in an advantageous embodiment of the method according to the invention, entropy coding of reconstruction levels is carried out. Quantization with maximization of mutual information results in reconstruction levels that are used with different probability. By entropy coding the reconstruction levels are assigned level identifiers or bit patterns of different lengths. A prefix-free code is used and more frequent reconstruction levels receive a shorter level descriptor. For example, a Huffman coding is used as entropy coding. As a result, a data compression is achieved with to be transmitted reconstruction levels. This leads to an effective reduction of a data rate when transmitting quantized likelihood quotients.
Eine vorteilhafte Ausbildung der erfindungsgemäßen Quantisierungseinrichtung bei einer Signalverarbeitung zum Quantisierung von Likelihood-Quotienten wird dadurch erreicht, dass Approximationsmittel zum Anpassen einer Näherungsfunktion als Funktion der Varianz an ermittelte Rekonstruktionsniveaus oder Entscheidungsniveaus bei verschiedenen Varianzen und ein Niveaugenerator zum Ermitteln von Rekonstruktionsniveaus und Entscheidungsniveaus bei einer Quantisierung mit der Näherungsfunktion vorgesehen sind. Wie bei der korrespondierenden Ausbildung des erfindungsgemäßen Verfahrens wird hierdurch eine zeit- und ressourcenaufwendige Maximierung der gegenseitigen Information durch ein Optimierungsverfahren vermieden. Optimale Niveaus in Abhängigkeit einer vorgegebenen Varianz werden schnell und unkompliziert ermittelt.An advantageous embodiment of the quantization device according to the invention in a signal processing for quantizing likelihood quotients is achieved in that approximation means for fitting an approximation function as a function of the variance to determined reconstruction levels or decision levels at different variances and a level generator for determining reconstruction levels and decision levels in a quantization with the approximation function are provided. As in the case of the corresponding embodiment of the method according to the invention, a time-consuming and resource-intensive maximization of the mutual information by an optimization method is thereby avoided. Optimal levels depending on a given variance are determined quickly and easily.
Weitere Ausgestaltungen der erfindungsgemäßen Quantisierungseinrichtung korrespondieren jeweils mit einer bereits beschriebenen Ausbildung des erfindungsgemäßen Verfahrens und verfügen über entsprechende Vorteile, die sichFurther embodiments of the quantization device according to the invention correspond in each case to an already described embodiment of the method according to the invention and have corresponding advantages that arise
aus dem Gegenstand der Unteransprüche ergeben.result from the subject of the dependent claims.
Ein Ausführungsbeispiel der Erfindung ist nachstehend unter Bezugnahme auf die zugehörigen Zeichnungen näher erläutert.An embodiment of the invention is explained below with reference to the accompanying drawings.
- Fig. 1Fig. 1
- zeigt in einem schematischen Flussdiagramm ein Ausführungsbeispiel des erfindungsgemäßen Verfahrens.shows a schematic flowchart of an embodiment of the method according to the invention.
- Fig. 2Fig. 2
- zeigt einen Vergleich von Rekonstruktionsniveaus ri und Entscheidungsniveaus di für R = 5 Rekonstruktionsniveaus, welche nach einem Verfahren zur Maximierung der gegenseitigen Information, einem direkt auf die L-Werte angewandtem Max-Lloyd-Verfahren und einem auf "Soft-Bits" der L-Werte angewandtem Max-Lloyd-Verfahren optimiert sind.Figure 4 shows a comparison of reconstruction levels r i and decision levels d i for R = 5 reconstruction levels, which, according to a method for maximizing mutual information, directly to the L values Max Lloyd method applied and a Max Lloyd method applied to "soft bits" of the L values.
- Fig. 3Fig. 3
- zeigt einen Vergleich des Verlusts an gegenseitiger Information Δ I zwischen quantisierten L-Werten und nicht quantisierten L-Werten als Funktion der mittleren gegenseitigen Information I C für R = 2 - 8 Rekonstruktionsniveaus. Dabei werden Quantisierungen mit Maximierung der gegenseitigen Information mit einer direkten Quantisierung der L-Werte und einer Quantisierung von "Soft-Bits" der L-Werte jeweils nach einem Lloyd-Max-Verfahren verglichen.FIG. 12 shows a comparison of the loss of mutual information Δ I between quantized L values and non-quantized L values as a function of the average mutual information I C for R = 2-8 reconstruction levels. In this case, quantizations with maximization of the mutual information are compared with a direct quantization of the L values and a quantization of "soft bits" of the L values in each case according to a Lloyd-Max method.
- Fig. 4Fig. 4
-
zeigt eine Tabelle mit angepassten Koeffizienten a, b und c der Näherungsfunktion
- Fig. 5Fig. 5
- zeigt Bitfehlerraten einer Log-MAP-Dekodierung eines PCC-Codes (G = [1,5/7], Blocklänge N = 2000, AWGN-Kanal) bei Verwendung einer Quantisierung durch Maximierung der gegenseitigen Information im Vergleich mit einer Verwendung einer direkten L-Wert-Quantisierung nach Lloyd-Max.shows bit error rates of a log MAP decoding of a PCC code ( G = [1.5 / 7], block length N = 2000, AWGN channel) using quantization by maximizing the mutual information compared to using a direct L - Value quantization according to Lloyd-Max.
- Fig. 6Fig. 6
- zeigt die Entropie von L-Werten, welche mit Maximierung der gegenseitigen Information Lq (X;L) quantisiert sind, als eine Funktion von σ2 L (durchgezogene Linien) und eine mit einer drei-dimensionalen Huffman-Kodierung erreichbare mittlere Bezeichnerlänge (Stufen).FIG. 12 shows the entropy of L values quantized with maximization of the mutual information L q ( X ; L ) as a function of σ 2 L (solid lines) and a mean identifier length achievable with a three-dimensional Huffman coding (stages ).
Im Folgenden werden zunächst mathematische Grundlagen und Ausdrücke näher erklärt und anschließend ein beispielhaftes Verfahren zur Quantisierung von Likelihood-Quotienten zusammen mit einer korrespondierenden Quantisierungseinrichtung beschrieben.In the following, first mathematical foundations and expressions are explained in more detail and then an exemplary method for the quantization of likelihood quotients described together with a corresponding quantization device.
Gegenseitige Information oder Transinformation ist eine dem Fachmann bekannte Größe aus der Informationstheorie, die die Stärke des statistischen Zusammenhangs zweier Zufallsgrößen angibt. Die gegenseitige Information 1(X; L) zwischen zwei Variablen x und l beschreibt die mittlere Reduzierung der Unsicherheit von x, wenn l bekannt ist und umgekehrt. Dieses wird beispielsweise in
Für einen gaußschen Kanal ist die Integration für den quantisierten, diskreten Fall durchführbar. Die gegenseitige Information zwischen X und dem zugehörigen quantisierten Log-Likelihood-Quotienten kann mit ausgeschriebener Summe über beide möglichen Bitwerte des Informationsbits X ∈ {+1, -1} dargestellt werden als:
Statt der Fehlerfunktion erf(x) kann auch die Standardnormalverteilung verwendet werden. Mit "ld" wird der Logarithmus zur Basis 2 bezeichnet. Mit
Die Differenz Δ I = Ic - I q zwischen den beiden gegenseitigen Informationen stellt den Informationsverlust durch die Quantisierung dar. Die Ableitungen nach den Rekonstruktionsniveaus ri für 1 ≤ i ≤ R lauten:
Die Ableitungen nach den Entscheidungsniveaus di für 1 ≤ i ≤ R+1 1 lauten:
Zur Minimierung des Informationsverlusts Δ I muss die gegenseitige Information I q der quantisierten L-Wert-Dichte maximiert werden. Bei einer vorgegebenen Varianz
In den Diagrammen der
Deutlich erkennbar ist in
In
In einem weiteren Verfahrensschritt, Rechteck 12, wird eine geeignete Näherungsfunktion der Varianz
In der Tabelle der
In einer alternativen Ausführung ist statt der Näherungsfunktion f(σ2) mit entsprechenden Koeffizienten eine Verwendung einer Wertetabelle für alle Rekonstruktions- und Entscheidungsniveaus für möglichst viele verschiedene Varianzen
Vor einer Quantisierung von L-Werten wird zunächst die Varianz
Im Folgenden werden diese Verfahrensschritte beispielhaft anhand einer Turbo-Decodierung eines parallelen verketteten Faltungscodes (PCCC: parallel concatenated convolutional code) verdeutlicht. Dabei wird zum Vergleich auch eine direkte Quantisierung von L-Werten nach Lloyd-Max verwendet und der Einfluss eines Quantisierungsverfahrens für L-Werte auf einen Signalverarbeitungsprozess aufgezeigt.In the following, these method steps are exemplified by means of a turbo-decoding of a parallel concatenated convolutional code (PCCC). For comparison, a direct quantization of L values according to Lloyd-Max is used and the influence of a quantization method for L values on a signal processing process is shown.
Als PCC-Code wird ein einfacher rekursiver systematischer PCCC mit Generatoren G = [1,5/7], Blocklänge N = 2000 und einem Zufalls-Interleaver mit Punktierung zum erzielen einer Rate ½ verwendet. Ein solcher Code ist dem Fachmann beispielweise aus
Zunächst wird die Varianz der L-Werte adaptiv an dem MAP-Decoderausgang mit dem Erfassungsmittel ermittelt. Dabei wird die gemessene Varianz von |L(X)|, den absoluten L-Wert der Codebits, auf die Varianz
Zum Vergleich werden Decodierungen mit unterschiedlichen R (Anzahl der Rekonstruktionsniveaus) bei der Quantisierung durchgeführt und statt der Quantisierung mittels Maximierung der gegenseitigen Information auch eine direkte Quantisierung der L-Werte nach dem Lloyd-Max-Verfahren vorgenommen. In
In einem weiteren Ausführungsbeispiel wird zusätzlich eine Entropiekodierung, Rechteck 20, der Rekonstruktionsniveaus mittels eines Entropiekodierers der Quantisierungseinrichtung durchgeführt. Entropiekodierungen sind dem Fachmann beispielsweise aus
Eine untere Grenze der notwendigen mittleren Bezeichnerlänge ist durch die Entropie der Rekonstruktionsniveaus gegeben:
In
Claims (10)
- A method in signal processing for quantizing likelihood quotients of binary random variables, wherein said quantized likelihood quotients are transmitted between two or more signal processing units,
characterized by
maximizing (10) the transinformation between the binary random variables and associated quantized likelihood quotients with a predetermined likelihood distribution of the likelihood quotients by variation of quantizing intervals and quantizing values, and
determining the quantizing interval and the quantizing values of maximum transinformation, and
quantizing (18) the likelihood quotients by using the quantizing interval and the quantizing values of maximum transinformation. - The method in signal processing for quantizing likelihood quotients pursuant to claim 1,
characterized by
collecting (14) a variance of the likelihood quotients prior to the quantizing for determining said quantizing intervals and quantizing values. - The method in signal processing for quantizing likelihood quotients pursuant to claim 2,
characterized by adapting (12) an approximation function as a function of the variance of said likelihood quotients to the determined quantizing values or quantizing intervals with different variances of said likelihood quotients, and determining (16) the quantizing values and quantizing intervals in quantizing with the approximation function. - The method in signal processing for quantizing likelihood quotients pursuant to claims 1 to 3,
characterized by using a value table and/or and interpolation when determining said quantizing values or quantizing intervals. - The method for quantizing likelihood quotients pursuant to claims 1 to 4, characterized in that entropy coding (20) of the quantizing values is performed.
- A quantizing device in signal processing for quantizing likelihood quotients of binary random variables, wherein said quantized likelihood quotients are transmitted between two or more signal processing units,
characterized in that
means for maximizing (10) the transinformation between the binary variables and the associated quantized likelihood quotients with a predetermined likelihood density of the likelihood quotients by variation of quantizing intervals and quantizing values, and
means for determining the quantizing interval and the quantizing values of maximum transinformation, and
means for quantizing (18) the likelihood quotients by using the quantizing interval and the quantizing values of maximum transinformation are provided. - The quantizing device in signal processing for quantizing likelihood quotients pursuant to claim 6,
characterized in that a collecting unit for collecting a variance of the likelihood quotients prior to the quantizing for the determining of said quantizing intervals and quantizing values is provided. - The quantizing device in signal processing for quantizing likelihood quotients pursuant to claim 7,
characterized by approximation means for adapting an approximation function as a function of the variance to the determined quantizing intervals or quantizing values with different variances of said likelihood quotients and a generator for determining the quantizing values and quantizing intervals in quantizing with the approximation function. - The quantizing device in signal processing for quantizing likelihood quotients pursuant to any of claims 6 to 8,
characterized in that the quantizing device comprises a memory with a value table and/or an interpolation unit for determining the quantizing values or quantizing intervals. - The quantizing device in signal processing for quantizing likelihood quotients pursuant to any of claims 6 to 9,
characterized in that an entropy encoder for the quantizing values is provided.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP09764705A EP2332301B1 (en) | 2008-09-26 | 2009-09-25 | Method and device for quantizing likelihood quotients |
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP08165233 | 2008-09-26 | ||
EP09764705A EP2332301B1 (en) | 2008-09-26 | 2009-09-25 | Method and device for quantizing likelihood quotients |
PCT/DE2009/075053 WO2010034309A2 (en) | 2008-09-26 | 2009-09-25 | Method and device for quantizing likelihood quotients |
Publications (2)
Publication Number | Publication Date |
---|---|
EP2332301A2 EP2332301A2 (en) | 2011-06-15 |
EP2332301B1 true EP2332301B1 (en) | 2013-03-13 |
Family
ID=42060157
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP09764705A Active EP2332301B1 (en) | 2008-09-26 | 2009-09-25 | Method and device for quantizing likelihood quotients |
Country Status (4)
Country | Link |
---|---|
US (1) | US8798174B2 (en) |
EP (1) | EP2332301B1 (en) |
ES (1) | ES2416065T3 (en) |
WO (1) | WO2010034309A2 (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11283539B2 (en) * | 2016-12-05 | 2022-03-22 | Telefonaktiebolaget Lm Ericsson (Publ) | Methods, encoder and decoder for handling a data stream for transmission between a remote unit and a base unit of a base station system |
US10069519B1 (en) * | 2018-01-23 | 2018-09-04 | Mitsubishi Electric Research Laboratories, Inc. | Partition based distribution matcher for probabilistic constellation shaping |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7299089B2 (en) * | 2004-08-20 | 2007-11-20 | Duke University | Methods, systems, and computer program products for neural channel selection in a multi-channel system |
US8243756B2 (en) * | 2007-07-12 | 2012-08-14 | Nokia Corporation | Methods, computer program products and apparatus providing improved quantization |
EP2272194B1 (en) * | 2008-04-25 | 2018-09-19 | Avago Technologies General IP (Singapore) Pte. Ltd. | Method and system for predicting channel quality index (cqi) values for maximum likelihood (ml) detection in a 2x2 multiple input multiple output (mimo) wireless system |
EP2134017B1 (en) * | 2008-05-09 | 2015-04-01 | Vodafone Holding GmbH | Method and system for data communication |
-
2009
- 2009-09-25 EP EP09764705A patent/EP2332301B1/en active Active
- 2009-09-25 ES ES09764705T patent/ES2416065T3/en active Active
- 2009-09-25 WO PCT/DE2009/075053 patent/WO2010034309A2/en active Application Filing
- 2009-09-25 US US13/121,360 patent/US8798174B2/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
ES2416065T3 (en) | 2013-07-30 |
US8798174B2 (en) | 2014-08-05 |
EP2332301A2 (en) | 2011-06-15 |
WO2010034309A3 (en) | 2010-11-11 |
US20110206166A1 (en) | 2011-08-25 |
WO2010034309A2 (en) | 2010-04-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
DE69718395T2 (en) | Information bit transmission method with error correction coding, coding and decoding device therefor | |
DE69412570T2 (en) | Iterative decoding method of chained block codes | |
DE602005003767T2 (en) | METHOD FOR COMPRESSING A LOT OF CORRELED SIGNALS | |
DE60316616T2 (en) | METHOD AND SYSTEM FOR CALCULATING THE BIT ERROR RATE OF A RECEIVED SIGNAL | |
DE60107527T2 (en) | Apparatus and method for halting iterative decoding in a CDMA mobile communication system | |
DE69936908T2 (en) | ITERATIVE DECODING OF PRODUCT CODES | |
DE602004001548T2 (en) | Method and apparatus for decoding a Low Density Partity Check (LDPC) code in a communication system | |
DE60217522T2 (en) | IMPROVED METHOD FOR CHARGING BIT ERRORS IN LANGUAGE CODING | |
DE20321883U1 (en) | Computer apparatus and system for entropy decoding quantized transform coefficients of a block | |
DE19526416A1 (en) | Method and arrangement for determining an adaptive termination criterion in the iterative decoding of multidimensionally coded information | |
DE102017130591B4 (en) | Method and device for error correction coding based on data compression | |
DE602004012417T2 (en) | DECODING DEVICE AND DECODING METHOD | |
EP0737389B1 (en) | Transmission system with soft output decoding and reduced storage capacity requirement | |
DE69107511T2 (en) | Coding method and coding arrangement with a subband encoder and transmitter with the coding arrangement. | |
DE60104338T2 (en) | DEVICE AND METHOD FOR TURBO DECODING WITH SIGNATURE DATA COMPARISON AS ABORT CRITERION | |
EP2332301B1 (en) | Method and device for quantizing likelihood quotients | |
DE60307429T2 (en) | Method and device for digital-analogue-to-analog-digital conversion with reduced energy consumption | |
DE60319557T2 (en) | Method for reducing the number of bits per soft bit | |
EP1133063A1 (en) | Method and device for source-controlled channel decoding wih a Kalman filter | |
DE112010002685B4 (en) | Method and device for decoding signals that are transmitted over fading channels with multiplying noise | |
EP1046254B1 (en) | Method and device for coding and transmitting information, using source-controlled channel decoding | |
DE19907728A1 (en) | Device and method for generating a data stream and device and method for reading a data stream | |
DE102011056132B4 (en) | Method and device for decoding data | |
EP1491015B1 (en) | Device and method for the controlled scaling and quantization of soft output values of an equalizer | |
DE19725275C2 (en) | Method for decoding block or convolutionally coded digital signals |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
17P | Request for examination filed |
Effective date: 20110420 |
|
AK | Designated contracting states |
Kind code of ref document: A2 Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO SE SI SK SM TR |
|
AX | Request for extension of the european patent |
Extension state: AL BA RS |
|
DAX | Request for extension of the european patent (deleted) | ||
17Q | First examination report despatched |
Effective date: 20120203 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R079 Ref document number: 502009006484 Country of ref document: DE Free format text: PREVIOUS MAIN CLASS: H04L0025030000 Ipc: H04L0025060000 |
|
GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
RIC1 | Information provided on ipc code assigned before grant |
Ipc: H04L 1/00 20060101ALI20120831BHEP Ipc: H04L 25/06 20060101AFI20120831BHEP |
|
GRAS | Grant fee paid |
Free format text: ORIGINAL CODE: EPIDOSNIGR3 |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO SE SI SK SM TR |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: FG4D Free format text: NOT ENGLISH |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: EP Ref country code: AT Ref legal event code: REF Ref document number: 601370 Country of ref document: AT Kind code of ref document: T Effective date: 20130315 |
|
REG | Reference to a national code |
Ref country code: IE Ref legal event code: FG4D Free format text: LANGUAGE OF EP DOCUMENT: GERMAN |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R096 Ref document number: 502009006484 Country of ref document: DE Effective date: 20130508 |
|
REG | Reference to a national code |
Ref country code: ES Ref legal event code: FG2A Ref document number: 2416065 Country of ref document: ES Kind code of ref document: T3 Effective date: 20130730 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: NO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130613 Ref country code: SE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: BG Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130613 Ref country code: LT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R082 Ref document number: 502009006484 Country of ref document: DE Representative=s name: KNH PATENTANWAELTE KAHLHOEFER NEUMANN ROESSLER, DE Ref country code: DE Ref legal event code: R082 Ref document number: 502009006484 Country of ref document: DE Representative=s name: KNH PATENTANWAELTE NEUMANN HEINE TARUTTIS PART, DE |
|
REG | Reference to a national code |
Ref country code: NL Ref legal event code: T3 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R082 Ref document number: 502009006484 Country of ref document: DE Representative=s name: KNH PATENTANWAELTE KAHLHOEFER NEUMANN ROESSLER, DE Ref country code: DE Ref legal event code: R082 Ref document number: 502009006484 Country of ref document: DE Representative=s name: KNH PATENTANWAELTE NEUMANN HEINE TARUTTIS PART, DE |
|
REG | Reference to a national code |
Ref country code: LT Ref legal event code: MG4D |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LV Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: GR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130614 Ref country code: FI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: SI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
REG | Reference to a national code |
Ref country code: NL Ref legal event code: T3 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: HR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: CZ Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: PT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130715 Ref country code: RO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: SK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: IS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130713 Ref country code: EE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: PL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: CY Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: DK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
26N | No opposition filed |
Effective date: 20131216 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R097 Ref document number: 502009006484 Country of ref document: DE Effective date: 20131216 |
|
BERE | Be: lapsed |
Owner name: VODAFONE HOLDING G.M.B.H. Effective date: 20130930 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MC Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: PL |
|
REG | Reference to a national code |
Ref country code: IE Ref legal event code: MM4A |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: BE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20130930 Ref country code: CH Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20130930 Ref country code: IE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20130925 Ref country code: LI Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20130930 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: SM Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: TR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: MT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LU Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20130925 Ref country code: MK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20130313 Ref country code: HU Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT; INVALID AB INITIO Effective date: 20090925 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: ES Payment date: 20150928 Year of fee payment: 7 |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: MM01 Ref document number: 601370 Country of ref document: AT Kind code of ref document: T Effective date: 20140925 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: IT Payment date: 20150924 Year of fee payment: 7 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: AT Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20140925 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: NL Payment date: 20150917 Year of fee payment: 7 |
|
REG | Reference to a national code |
Ref country code: FR Ref legal event code: PLFP Year of fee payment: 8 |
|
REG | Reference to a national code |
Ref country code: NL Ref legal event code: MM Effective date: 20161001 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: NL Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20161001 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: IT Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20160925 |
|
REG | Reference to a national code |
Ref country code: FR Ref legal event code: PLFP Year of fee payment: 9 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: ES Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20160926 |
|
REG | Reference to a national code |
Ref country code: FR Ref legal event code: PLFP Year of fee payment: 10 |
|
REG | Reference to a national code |
Ref country code: ES Ref legal event code: FD2A Effective date: 20181120 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20230920 Year of fee payment: 15 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: FR Payment date: 20230928 Year of fee payment: 15 Ref country code: DE Payment date: 20230920 Year of fee payment: 15 |